This was the first time I used a ‘text to image’ AI
generator, and I found the process both interesting and wildly frustrating. I
have never been a visual artist- I cannot draw, and my overall visual
creativity has always been lackluster. My strength has always been as a writer,
so I found it fascinating to turn my words into art that is well beyond my
skill level to create by hand. The frustration occurred in finding the correct
descriptors and word combinations to transfer what was in my head to the AI
generator.
Being unfamiliar with how an AI Generator works, I first
tried an ultra-detailed description right down to the artist whose style I
wanted the image to emulate. That resulted in my first problem, as it seems
there is a block or ban (at least with the generator I used) on using specific,
well-known individuals as templates for the generator. Once I removed the
(copyrighted?) names, it started producing images that were nothing like what I
visualized. After some experimentation, I started simplifying my prompts and
removing specific, real-life visual references. This was more successful, but I
still struggled to get what I wanted. I used 51 different prompt
combinations, resulting in 204 pictures.
In the end, I settled on these AI selfies because there are
elements of each I like, but I was never able to figure out the correct
combination of words to produce a single picture. One of the biggest problems I
had with my earliest AI selfies was that each turned me into a contestant on
America’s Next Top Model. I don’t have that self-perception of myself, so I had
to work to tone down the Zoolander a little bit.
I love comics and running, so that added some personal flair.
And even though I kind of look like an amalgamation of young Clark Kent and
Peter Parker from a retro comic- it’s a lot better than smoldering supermodel
Brandon.
I tried to convey the sense of someone a little bit out of
touch with modern digital technology with the contrast between the retro comic
style and modern platforms. I like the avatar in the first picture more and
tried to combine it with the digital technology Big Bang background of the
other selfie, but it never quite came to fruition.
Also, I didn't achieve a word combination that created a pictorial representation of AI without it devolving into stereotype (e.g.,
evil-looking robots). All in all, for a first foray into AI digital selfies,
I’m happy with how it turned out.
Comments
Post a Comment